Multi class support vector machine implementation to intrusion detection

Despite advances in security practices the threat to information assurance is on the rise. Due to the growing number of malicious usage, attacks, stealing of sensitive information and sabotage, information security has become one of the prime concerns for many governments as well as corporate organizations, the world over. There exists a constant need for improvement and innovation in detection of intrusions and adoption of efficient countermeasures against security breaches. In a new approach, this paper focuses on applying multi class support vector machine classifiers, using one-versus-one method, for anomalous as well as misuse detection to identify attacks precisely by type. Evaluation has been done over a benchmark dataset used in the Third Knowledge Discovery and Data mining competition (KDD'99). The results obtained are comparable to some of the best in the contest.